1. Introduction

Evergreen broad-leaved forests (EBLFs) are a forest community type consisting of evergreen dicotyledonous plants with high endemism and rich biodiversity, dominated by families such as Fagaceae, Lauraceae, Magnoliaceae and Theaceae (Wu, 1995; Song, 2013; Tang, 2015; Song and Da, 2016; Fig. 1). These forests occupy large areas in East Asia, especially in southern China, southern Japan, and southernmost Korea (Song and Da, 2016; Fig. 1). Compared to other latitudinal bands worldwide, particularly those with semi-arid or arid desert conditions, EBLFs predominantly occur under monsoon climates and exhibit significantly higher productivity in East Aisa (Tang, 2015; Song and Da, 2016). They hold significant importance as terrestrial ecosystems, substantially enhancing global forest ecosystem services, providing various biological resources, and promoting sustainable social development (Piao et al., 2009; Pan et al., 2011; Fang et al., 2014; Tang, 2015). Therefore, understanding the historical dynamics of EBLFs and their underlying drivers is crucial for conserving their diversity in East Asia. However, the origin and evolution of EBLFs remains controversial across different disciplines.

The historical dynamics of EBLFs have been reconstructed using molecular approaches. For instance, molecular dating of Fagaceae (Xiang et al., 2014; Xing et al., 2014; Hipp et al., 2020; Hai et al., 2022), Theaceae (Yu et al., 2017; Rao et al., 2018), Lauraceae (Li et al., 2011; Huang et al., 2016a; Xiao et al., 2022; Qin et al., 2023), and Magnoliaceae (Azuma et al., 2001; Nie et al., 2008; Dong et al., 2021) have all indicated that EBLFs emerged in the Eocene. Other key groups, e.g., Dendrobium (Orchidaceae), Quercus Section Cyclobalanopsis (Fagaceae) and Mahonia (Berberidaceae), have suggested that EBLFs may have been established in mainland Asia during the early Oligocene (Xiang et al., 2016; Deng et al., 2018; Chen et al., 2020). Meta-analysis of 72 generic-level characteristic lineages has also demonstrated that the origin of EBLFs can be traced back to the early Oligocene (Zhang et al., 2024). Several synthetic studies using many lineages have proposed that the East Asian flora (EAF) might be relatively young, with most of its clades originating since the Miocene (Chen et al., 2018; Lu et al., 2018). Furthermore, many studies have demonstrated that a rapid radiation occurred in East Asia during the Miocene (Yu et al., 2017; Deng et al., 2018; Chen et al., 2020; Yan et al., 2021; Hai et al., 2022; Zhang et al., 2022).

Plant fossil records provide more direct and reliable evidence for the origin and evolution of EBLFs. High-quality geochronological studies have called for re-examining the emergence of these forests. The previously presumed Miocene strata (WGCPC, 1978; BGMRYP Bureau of Geology and Mineral Resources of Yunnan Province, 1990, BGMRYP Bureau of Geology and Mineral Resources of Yunnan Province, 1996; Xu et al., 2008; Li et al., 2015; Huang et al., 2016b, 2017, 2018) for many fossil floras in southwestern China has now been revised to be late Eocene/early Oligocene in age (Gourbet et al., 2017; Linnemann et al., 2018; Su et al., 2019; Tian et al., 2021). Evergreen representatives, including Fagaceae, Lauraceae and Fabaceae, are main components in these floras (Huang, 2017; Su et al., 2019; Deng et al., 2020; Tian et al., 2021; Wu et al., 2022b). The revised geological ages and components of fossil floras indicate that the temporal occurrence of EBLFs may have been advanced to the late Eocene–early Oligocene in southwestern China. Likewise, plant fossil records in southern China with high diversity of evergreen elements demonstrate that EBLFs emerged in this region at least from the middle Eocene (Spicer et al., 2014, 2016; Aleksandrova et al., 2015; Herman et al., 2017; Averianov et al., 2017). The discrepancy between molecular and paleobotanical evidence suggests a complex evolutionary history of EBLFs in East Asia, and a comprehensive understanding of this issue requires deep exploration.

Long-term climate changes have influenced the development of specific biomes over time. The Asian monsoon has played a critical role in the formation of EBLFs by bringing a humid climate that has facilitated their historical assembly (Song and Da, 2016; Yu et al., 2017; Li et al., 2021a). The relationship between the evolution of the Asian monsoon and the emergence of EBLFs remains unclear, probably because the timing of the onset of the Asian monsoon is controversial. Some researchers have suggested that the Asian monsoon originated around the Oligocene–Miocene boundary (Sun and Wang, 2005; Guo et al., 2008; Lin et al., 2015), whereas others argue that it emerged as early as the Eocene (Wang et al., 2013; Quan et al., 2014; Licht et al., 2014; Li et al., 2022) or even the Paleocene (Spicer et al., 2016; Spicer, 2017). A recent review that integrated proxy data and modeling evidence proposed a two-stage northward expansion process for the Asian monsoon (Wu et al., 2022a). This study suggests that the dynamics of EBLFs may undergo multiple stages, providing us an opportunity to deeply explore how climate variations influence the spatial and temporal patterns of vegetation.

In this study, we use paleobotanical evidence to identify when and where EBLFs have occurred in East Asia. Based on a large dataset of the Cenozoic macrofossil records in East Asia, we reconstructed the fossil history of EBLFs' dominant genera, determined the most similar extant vegetation, calculated the paleoclimate for each fossil assemblage, as well as simulated the paleoclimate during the Cenozoic. We further explored the relationship between the historical dynamics of EBLFs and climate evolution. Our results demonstrate that occurrence of EBLFs in East Asia is heterogeneous, which is closely linked with the evolution of the Asian Monsoon.

2. Materials and methods 2.1. Paleobotanical data

We compiled a dataset of Cenozoic paleobotanical records from East Asia, focusing on macrofossils such as leaves, seeds, fruits, and wood to minimize biases caused by potentially long-distance transport of pollen grains. We excluded fossil assemblages with fewer than ten taxa, resulting in a selection of 50 assemblages from 47 fossil floras (Fig. 1 and Table 1). Geographic locations and climatic conditions (Song, 2013; Song and Da, 2016) were used to classify the fossil assemblages into four regions: southwestern China (including southeastern Xizang and Yunnan), southern China (including Guangxi, Guangdong, and Hainan), central-eastern China (including Jiangxi, Shandong, Zhejiang, and Fujian), and Japan (including central and southwestern regions) (Fig. 1 and Table 1). More detailed information on each fossil assemblage is provided in Table S1. Moreover, we identified the dominant genera among all fossil taxa based on their roles in modern EBLFs (Table S2) and calculated the probability density distribution of these genera appearing in different geological periods within each region.

2.2. Flora similarity analysis

We used the Phytogeographic Reference Region Assessment (PRRA) approach to determine the most similar extant forest vegetation for each fossil assemblage (Kunzmann et al., 2022). Prior to analysis, the fossil taxa were assigned to their nearest living relatives (NLRs). The NLRs were primarily retrieved from primary literature; if NLRs were not mentioned in the literature, we referred to the PALAEOFLORA database (http://www.palaeoflora.de). Distribution data for NLRs were obtained from the Global Biodiversity Information Facility database (GBIF, http://www.gbif.org) and processed using the methods described by Palazzesi et al. (2014) and Kunzmann et al. (2022). By calculating the number of overlapping taxa between NLRs and modern global vegetation data to assess the degree of similarity in grids for each fossil assemblage. Then we calculated the mean similarity across the four regions over different geological periods. Given the high similarity between them, the modern vegetation type of high similarity grids can be considered the same as the vegetation type of fossil assemblages. PRRA analysis was conducted in R with a 2° grid resolution, following the procedure outlined by Kunzmann et al. (2022).

2.3. Quantitative reconstruction of the paleoclimate

We used Joint Probability Density Functions (JPDFs) to reconstruct the paleoclimate of each fossil assemblage (Willard et al., 2019). This method assumed that the ecological tolerance of plant fossils is comparable to their NLRs (Utescher et al., 2014). Thus, taxa that exhibit significant changes in their distribution over geological time, such as relict plants (e.g., Metasequoia, Ginkgo, Glyptostrobus, and Cyclocarya), should not be considered for reconstruction process. Several widespread fossil taxa were also excluded in the reconstruction process, as they would have been of limited value for determination of climatic conditions, such as Pinaceae, Poaceae, and Polygonaceae.

Similar to the PRRA approach described in section 2.2, the JPDFs also depend on the climatic distribution of NLRs. Raw data derived from GBIF were cleaned by removing repeated coordinates, erroneous coordinates, cultivation sites, wherever possible, and gridded (keep only one occurrence in 5′ × 5′ area) distribution data using the "gridSample" function of the R package "dismo". Modern climatic data were obtained from CHELSA (Climatologies at high resolution for the earth's land surface areas) climate database (http://chelsa-climate.org, Karger et al., 2017). The most likely climatic conditions of each fossil assemblage were then calculated by cross-plotting the probability density of each taxon (Willard et al., 2019; West et al., 2020).

In total, six climatic variables were calculated using this quantitative method, namely mean annual temperature (MAT, ℃), mean temperature of the warmest month (MTWM, ℃), mean temperature of the coldest month (MTCM, ℃), mean annual precipitation (MAP, mm), precipitation of the wettest quarter (PWetQ, mm), and precipitation of the driest quarter (PDryQ, mm).

2.4. Numerical simulations of the paleoclimate

We used the Hadley Centre Coupled Model (v.3 low resolution; HadCM3BL) from the University of Bristol in the UK, to simulate paleoclimatic conditions (Valdes et al., 2021; Fenton et al., 2023). The HadCM3BL is a fully coupled atmosphere-ocean general circulation model with a low resolution of 3.75° longitude × 2.5° latitude, 19 vertical atmosphere levels and 20 ocean depth levels (Valdes et al., 2017). The land surface scheme used the Met Office Surface Exchange Scheme (MOSES 2.1) and the vegetation predicted used a dynamic vegetation and terrestrial carbon cycle scheme, TRIFFID (Top-down Representation of Interactive Foliage and Flora Including Dynamics) (Cox, 2001; Valdes et al., 2017).

The boundary conditions were constrained as follows: the solar constant of each period was increased linearly to the modern value, and the earth orbit parameters were set to present conditions. The land-sea boundary, topography, and ice sheets were derived from the Scotese (2021). CO₂ concentrations were set according to the values presented in Foster et al. (2017) (Table S3). Based on these boundary conditions, the paleoclimate of different periods during the Cenozoic was simulated. Each snap shot experiment was run for 1000 model years and reached an equilibrium state in both the surface and deep ocean, with no energy imbalance at the top of the atmosphere (Valdes et al., 2021). The mean of the last 100 model years was regarded as the simulated outputs in our research. The raw output climate data were interpolated to high-resolution data (5′) to facilitate visualization in figures and comparison with fossil data using the "resample" function from the "terra" package in R. Numerical simulations offer a thorough representation of climatic patterns, providing valuable insights to supplement the limited temporal and spatial coverage of fossil records. As such, the simulated results were compared to reconstructions based on fossil records.

To evaluate the performance of the model, we compared the simulated modern output (Fig. 5I) with a very high resolution (30 arc sec, ~1 km) observed data in 1981~2010 from CHELSA climate database (http://chelsa-climate.org, Karger et al., 2017; Fig. S1). The simulated modern output was generally consistent with the observations. However, there was a slight discrepancy in northern China, which is likely due to the resolution difference between the model and observations. In addition, through a detailed comparison with other models available in Coupled Model Intercomparison Project Phase 5 (CMIP5), it has been shown that the HadCM3BL performed relatively well (Valdes et al., 2017). Therefore, this model can be used to capture the broad features of the climate, and provide important clues for understanding Cenozoic climate change.

3. Results 3.1. The fossil history of dominant genera for East Asian EBLFs

Probability density distribution analysis indicate that the timing of appearance for EBLFs' dominant genera varied across four regions (Fig. 2). In southern China, dominant genera appeared as early as the Paleocene–early Eocene and peaked in the middle Eocene, followed by a continuous decline (Fig. 2A). While they firstly appeared in southwestern China during the middle Eocene and peaked in the late Eocene–early Oligocene, after that dominant genera declined rapidly in the late Oligocene–early Miocene before rebounding in the late Miocene (Fig. 2B). In Japan, dominant genera appeared in the middle Eocene, peaked in the early Oligocene, and continued to decrease with a sub-peak in the late Miocene–Pliocene (Fig. 2C). In central-eastern China, the peak of dominant genera occurred around the middle Miocene. Afterwards, two consecutive small peaks occurred in the late Miocene and Pliocene (Fig. 2D).

3.2. Similarity analysis results for fossil assemblages

PRRA results showed that high similarity areas (the colored red grid boxes with high co-occurrence values) for the four examined regions differ, but each region nearly matches its modern location (Fig. 3). In southern China, the concentration of high similarity areas between fossil assemblages and modern vegetation generally increased from the middle Eocene to Miocene (Fig. 3A). During the middle Eocene, the high similarity area was widely distributed in southern China (Fig. 3A1), but then gradually contracted to the northern part of the Indochina Peninsula (Figs. 3A4 and S10). In southwestern China, high similarity areas have changed slightly and are mainly distributed in southeastern Xizang, Yunnan and Sichuan (Fig. 3B). In comparison to other periods, the similarity area was more concentrated in northwestern Yunnan during the Pliocene (Fig. 3B4). In the early Oligocene, the high similarity area of the Markam assemblage was not only distributed in southwestern China, but also in southwestern Japan and western Europe (Fig. S12). In Japan, high similarity areas have undergone significant changes. These areas were originally located in southeastern China during the late Eocene (Fig. 3C1). However, they shifted to southwestern Japan from the Oligocene (Fig. 3C2–C4). In central-eastern China, high similarity areas were widely distributed in eastern China, with the highest concentration in Jiangxi and Anhui (Fig. 3D). The similarity areas of middle Miocene Zhangpu assemblage were mainly distributed in Southeast Asia (Fig. S29).

3.3. Quantitative paleoclimate reconstruction from plant fossils

Table S4 shows the paleoclimatic data using the Joint Probability Density Functions (JPDFs) for each plant fossil assemblage of East Asia, including six climate variables, i.e., MAT, MTCM, MTWM, MAP, PWetQ, PDryQ. Of these, MAT, MTCM and MAP were employed to assess whether a fossil assemblage fulfills the climatic criteria of EBLFs in East Asia.

The EBLFs on the East Asian occurred under monsoon climate, with hot, moist summers and fairly cold winters. Our reconstructed paleoclimates show that EBLFs in each regions exhibited distinct climatic conditions. EBLFs cannot be established if dominant taxa are unable to tolerate climatic conditions of each region (the climatic conditions shown by dashed lines in Fig. 4). The key limiting factor for the establishment of EBLFs was MTCM, with values needing to exceed the minimum threshold to prevent damage from freezing temperatures.

In southern China, the emergence of EBLFs requires the highest temperature and precipitation. Our reconstruction indicated that MAT ranged from 19 ℃ to 21 ℃, MTCM from 9 ℃ to 12 ℃, and MAP from 1500 mm to 2200 mm (shown by dashed line in Fig. 4A). Most of the reconstructed paleoclimates align with these climatic conditions. However, the reconstructed MAT of the late Eocene Maoming assemblage was below the minimum limit, ranging from 12.95 ℃ to 17.35 ℃ (Fig. 4A1, No. 5). Conversely, the reconstructed MTCM of the Miocene Guiping assemblage exceeded the maximum limit, ranging from 13.15 ℃ to 15.95 ℃ (Fig. 4A2, No.9).

The climatic conditions conducive to the emergence of EBLFs in the southwestern China were relatively lower, with MAT ranging from 10 ℃ to 20 ℃, MTCM from 4 ℃ to 12 ℃, and MAP from 800 mm to 1500 mm (shown by dashed line in Fig. 4B). The reconstructed MAT and MAP of all fossil assemblages in this region satisfy these criteria, except for the early Oligocene Markam assemblage, where the reconstructed MTCM falls below the minimum threshold, ranging from −4.15 ℃ to 3.95 ℃ (Fig. 4B2, No.11).

EBLFs occurred in Japan under climatic conditions that required lower temperatures and higher precipitation. The MAT ranged from 13 ℃ to 18 ℃, MTCM from 0.9 ℃ to 10 ℃, and MAP from 1200 mm to 2500 mm (shown by dashed line in Fig. 4C). The reconstructed MTCM and MAP of all fossil assemblages in this region meet these conditions, except the reconstructed MAT of the early Miocene Shichiku assemblage, which is lower than the minimum limit, ranging from 9.85 ℃ to 12.85 ℃ (Fig. 4C1, No.40).

Compared with other regions, the climatic conditions for the emergence of EBLFs in central-eastern China were relatively moderate, with MAT ranging from 15 ℃ to 18 ℃, MTCM from 2 ℃ to 10 ℃, and MAP from 1000 mm to 2000 mm (shown by dashed line in Fig. 4D). The reconstructed climates of most fossil assemblages in this region align with these conditions, except for the middle Miocene Zhangpu assemblage, which exceeds the maximum standard, with reconstructed MAT and MTCM ranging from 20.55 ℃ to 24.15 ℃ and 16.65 ℃–21.95 ℃, respectively (Fig. 4D1 and D2, No.28).

3.4. Comparing paleoclimates from quantitative reconstruction and model simulation

Compared to other climatic factors (Figs. S52–S56), we observed a strong correspondence in PWetQ between the quantitative reconstructions from plant fossils and model simulations, especially in regions where PWetQ exceeded 600 mm (Fig. 5). However, there are some exceptions, for example, the quantitatively reconstructed PWetQ for the late Eocene Jianchuan and Markam assemblages (Fig. 5B), as well as middle Miocene Shanwang assemblage (Fig. 5F), which exceeds simulation results. In contrast, for some Miocene–Pliocene fossil assemblages, reconstructed values are lower than those simulated by models.

Simulation results indicate that PWetQ varied in response to paleogeographical changes (Fig. 5). In the early Eocene, southeastern China experienced abundant precipitation (Fig. 5A), whereas it became relatively dry in the late Eocene (Fig. 5B). During the Oligocene, precipitation in southwestern China gradually increased as a result of the continued uplift of the Tibetan Plateau, leading to a reduction in the central arid zone (Fig. 5C and D). Since the Miocene, southern China has generally witnessed increased precipitation, resulting in the disappearance of the arid zone (Fig. 5E–I).

4. Discussion 4.1. How to define EBLFs by plant fossils

One central question is how fossil plants can be used to identify historical occurrences of EBLFs. In this study, we use the components, paleoclimate, and paleovegetation of plant fossil assemblages to identify these historical occurrences of EBLFs.

EBLFs can be defined by the presence of EBLFs' components in fossil assemblages, e.g., Fagaceae, Lauraceae, Theaceae, and Magnoliaceae. These components are dominant in modern EBLFs and, therefore, serve as important indicators that identify this biome (Song, 2013; Song and Da, 2016). Over geological time, plant fossil assemblages exhibit distinct components and reflect various vegetation types. Some fossil assemblages (e.g., the late Eocene Jianchuan assemblage and the early Oligocene Wenshan assemblage in southwestern China, middle Eocene Changchang and Maoming assemblages in southern China, late Eocene Ube assemblage in Japan, late Miocne Tiantai flora in central-eastern China) represent evergreen broadleaved forests, of which Fagaceae and Lauraceae are the principal components, while Hamamelidaceae, Fabaceae and other evergreen groups are also common (Huzioka and Takahasi, 1970; WGCPC, 1978; Li, 2010; Spicer et al., 2014, 2016; Aleksandrova et al., 2015; Herman et al., 2017; Huang, 2017). Some fossil assemblages (e.g., Pliocene Yongping, Lanping and Eryuan assemblages in southwestern China) represent evergreen sclerophyllous broad-leaved forests, with Quercus sect. Heterobalanus and conifers (Pinus, Abies, Picea, Tsuga) as the dominant elements (Tao and Kong, 1973; Huang et al., 2012, 2013, Huang et al., 2015, Huang et al., 2020; Su et al., 2013; Zhu et al., 2015, 2016). Some fossil assemblages (e.g., late Eocene Markam assemblage and early Oligocene Lühe assemblage in southwestern China, late Oligocene Noda assemblage in Japan and middle Miocene Shanwang assemblage in central-eastern China) represent evergreen-deciduous broadleaved mixed forests, dominated by the evergreen groups of Fagaceae, Magnoliaceae, Lauraceae and the deciduous groups of Betulaceae, Ulmaceae, Salicaceae (Hu and Chaney, 1940; Tanai and Uemura, 1991; Su et al., 2019; Deng et al., 2020; Wu et al., 2022b). Some fossil assemblages (e.g., early Oligocene Markam assemblage) represent alpine deciduous shrubs. Small-leaved shrubs such as Salix, Rosa and Quercus sect. Heterobalanus are main components (Su et al., 2019; Deng et al., 2020). Additionally, some fossil assemblages (e.g., middle Miocene Zhangpu assemblage and Guiping assemblage) represent tropical rainforests, dominated by tropical taxa including Annonaceae, Myrtaceae, and Burseraceae (Wang, 2018; Wang et al., 2021; Huang et al., 2021; Li et al., 2021b; Wu et al., 2021; Song et al., 2023).

Secondly, appropriate climatic conditions are essential for the emergence of EBLFs, as climate plays a key role in determining their geographic range. Our paleoclimatic analysis of plant fossils indicates that most fossil assemblages are satisfactory for the emergence of EBLFs, although some are not (Fig. 4). The reconstructed MTCM is lower in the early Oligocene Markam assemblage than today. This may be due to the uplift of southeastern Xizang and a cooling trend during the early Oligocene, leading to an increase in deciduous components (Su et al., 2019; Deng et al., 2020). The MAT reconstructed for the early Miocene Shichiku assemblage falls below the minimum threshold because both the composition and foliar physiognomy of this assemblage resemble the Aniai-type flora. The Aniai-type flora is characterized by cool-temperate vegetation consisting of deciduous broad-leaved trees and conifers (Tanai, 1961; Yabe, 2008). The reconstructed temperatures of middle Miocene Zhangpu assemblage and Guiping assemblage exceed the maximum values on account of the Mid-Miocene Climatic Optimum (MMCO, ~14–17 million years ago), which resulted in a warmer climate leading to the presence of many tropical components in them (Wang, 2018; Wang et al., 2021; Huang et al., 2021; Li et al., 2021b; Wu et al., 2021; Song et al., 2023). In addition, the NLRs of some macrofossils were only identified at the genus or even family levels, which could potentially restrict the effectiveness of the method employed for climate reconstruction (Liu et al., 2011). This limitation may account for the lower reconstructed MAT of the middle Eocene Maoming assemblage. In comparison, we observed a high level of consistency between the quantitative reconstruction and simulated outputs in regions where PWetQ exceeds 600 mm (Fig. 5). This finding is also verified by the distribution of modern EBLFs and PWetQ in current observations (Fig. S1). Therefore, we deduce that PWetQ exceeding 600 mm is an important factor for the occurrence of EBLFs.

Thirdly, we can determine if a fossil assemblage represents an EBLF from its most similar extant forest vegetation. This is possible because modern vegetation in high-value grids can be regarded as the vegetation type most closely resembling that of the fossil assemblage. According to the distribution range of modern EBLFs (the white dashed lines in Fig. 1), regions with high similarity to most fossil assemblages predominantly coincide with the main areas of East Asian EBLFs today. Consequently, the paleovegetation of most fossil assemblages can be classified as EBLFs. In addition, the paleovegetation of a few fossil assemblages does not exhibit features of EBLFs due to their distinct composition. For example, the middle Miocene Shanwang and Guiping assemblages contain many tropical taxa, leading to a high similarity with areas located at lower latitudes in Southeast Asia. These areas are currently characterized by tropical monsoon forests or rainforests.

With a few exceptions (e.g., middle Miocene Shanwang and Guiping assemblages, early Oligocene Markam assemblage), the components, paleoclimate, and paleovegetation of plant fossil assemblages were able to identify EBLFs in East Asia during geological time.

4.2. Heterogeneous occurrence of evergreen broadleaved forests in East Asia

Combined with the peak period of appearance for EBLFs' dominant genera (Fig. 2), we inferred that occurrences of EBLFs show different temporal–spatial patterns. The earliest occurrence was in southern China during the middle Eocene (Fig. 6A), followed by southwestern China in the late Eocene–early Oligocene (Fig. 6B and C), then Japan in the early Oligocene, and finally central-eastern China during the Miocene (Fig. 6C and D). Furthermore, our comparison of quantitative reconstruction and model simulation has led us to propose that PWetQ exceeding 600 mm may play a critical role in the occurrence of EBLFs (Fig. 6).

To some extent, our findings align with previous molecular dating studies (Xiang et al., 2016; Yu et al., 2017; Xiao et al., 2022; Qin et al., 2023; Zhang et al., 2024). Whether EBLFs originated in the Eocene, early Oligocene, or Oligocene–Miocene boundary, the paleobotanical evidence seems to support these estimates. However, evidence from molecular dating has primarily focused on the temporal emergence of EBLFs, overlooking their spatial distribution. The unique taphonomy of plant fossils allow for a direct exploration into the spatial occurrence of EBLFs in East Asia.

Moreover, when considering only paleobotanical evidence, our research indicates an earlier occurrence than previously thought, especially in southwestern China. This is due to the recent updates in the geological ages of the fossil assemblages collected in this region. This new chronological framework has advanced the occurrence time of EBLFs to the late Eocene–early Oligocene in southwestern China (Gourbet et al., 2017; Linnemann et al., 2018; Su et al., 2019; Tian et al., 2021).

4.3. The origin of evergreen broad-leaved forests in response to the evolution of the Asian Monsoon

The Asian Monsoon played a crucial role in the formation and development of EBLFs, which are especially impacted by precipitation (Song and Da, 2016; Yu et al., 2017; Li et al., 2021a; Ye and Li, 2022; Zhang et al., 2024). A recent review has demonstrated that the Asian Monsoon has undergone three phases of spatial expansion, indicating that EBLFs may have had a complex occurrence (Wu et al., 2022a).

The initiation of the Asian Monsoon is the tropical Asian Monsoon prevailling in southern Asia (the south of 20–22°N) before ~41 Ma, which was largely driven by the seasonal fluctuation of the Intertropical Convergence Zone (ITCZ) (Trenberth et al., 2000; Wang, 2009; Wang et al., 2014; Spicer et al., 2016; Farnsworth er al., 2019; Tardif et al., 2020). Previous studies have suggested that under the influence of tropical Asian Monsoon, the climate became humid and facilitated the appearance of EBLFs in southern China (Herman et al., 2017; Jin et al., 2017; Spicer, 2017). Palynological records across southern China have also observed a major environmental transition: from a relative high abundance of xerophilous taxa in the Paleocene–early Eocene to subtropical wet evergreen and deciduous broad-leaved mixed forest in middle Eocene (Xie et al., 2019, 2020). Our study also demonstrates that EBLFs occurred in southern China during the middle Eocene, providing further evidence to support this hypothesis (Fig. 6A).

Subsequently, as a consequence of the full collision of India with Asia, fast northward movement of the southern margin of the Tibetan Plateau and Antarctic cooling, the Asian Monsoon expanded northward to the southern subtropical region (~26°N) ~41 Ma, synchronously promoting an abrupt dry–wet transitions and providing a unique opportunity for the emergence of EBLFs in southwestern China (Fig. 6B; Fang et al., 2021).

Finally, the modern-like pattern of the Asian Monsoon was fully established at ~26 Ma, probably driven by the uplift of the Tibetan Plateau and global climate change (Wu et al., 2022a). The ultimate establishment of the Asia Monsoon has shaped "a north-west dry and south-east humid" climate pattern in East Asia, which has been recognized as a major contributor to the formation of EBLFs in central-eastern China and Japan (Fig. 6C and D; Sun and Wang, 2005; Guo et al., 2008). Recent research has integrated modeling results and fossil data to demonstrate that the growth of north and northeastern Xizang altered the Asian monsoon system in the late Oligocene. This change led to a transition from deciduous broadleaf vegetation to evergreen broadleaf vegetation and increased plant diversity across southeastern Asia (Li et al., 2021a).

5. Conclusion

By intergrating the latest advances in paleobotany, our research has explored the occurrence of the EBLFs in East Asia from paleobotanical evidence. Different from the previously prevailing views in historical assembly of EBLFs, our findings highlight that the EBLFs in East Asia appeared in different temporal–spatial patterns: the earliest occurrence was in southern China in the middle Eocene, followed by southwestern China in the late Eocene–early Oligocene, Japan during the early Oligocene, and central-eastern China around the Miocene. In addition, we propose that PWetQ exceeding 600 mm is an important factor for the formation of EBLFs. Combined with the latest studies on the Asian monsoon, our findings suggest that the multistage evolution of Asian monsoon led to the diverse occurrence of EBLFs in East Asia.

Acknowledgements

The author would like to express gratitude for the multiple warm invitations extended by the Ailao Mountain Ecology Station to visit and study at the site. It was while facing the sprawling and undulating evergreen broad-leaved forests that the author began to ponder the intriguing scientific question of the how and when of these forests have occurred. The author is also appreciative of the colleagues and students from the Paleoecology Research Group who participated in fieldwork, providing diligent and thorough contributions. We would like to thank Professor Jianhua Jin for sharing the information on paleobotany of southern China and for his valuable discussions. We also would like to thank Dr. Tammo Reichgelt for help with paleoclimatic reconstruction using plant fossils. This study was supported by National Key R&D Program of China (No.2022YFF0800800), National Science Fund for Distinguished Young Scholars (No. 32225005), National Natural Science Foundation of China (NSFC) (Nos. 42072024, 42320104005, 42372033), the Young and Middle-aged Academic and Technical Leaders of Yunnan (No.202305AC160051), Basic Research Project of Yunnan Province (No. 202401AT070222), the 14th Five-Year Plan of the Xishuangbanna Tropical Botanical Garden, Chinese Academy of Sciences (Nos. XTBG-1450101, E3ZKFF7B).

CRediT authorship contribution statement

Jiagang Zhao: Writing – review & editing, Writing – original draft, Visualization, Formal analysis, Data curation. Shufeng Li: Writing – review & editing, Methodology, Formal analysis. Jian Huang: Visualization, Formal analysis. Wenna Ding: Visualization, Formal analysis. Mengxiao Wu: Writing – review & editing, Formal analysis. Tao Su: Writing – review & editing, Visualization. Alexander Farnsworth: Visualization, Methodology. Paul J. Valdes: Visualization, Methodology. Linlin Chen: Writing – review & editing, Visualization, Methodology. Yaowu Xing: Writing – review & editing, Writing – original draft, Visualization, Funding acquisition. Zhekun Zhou: Writing – review & editing, Writing – original draft, Funding acquisition, Conceptualization.

Declaration of competing interest

The authors declare no conflict of interests.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2024.07.004.